Anda di halaman 1dari 17

SEISMIC ANALYSIS MODELING

TO SATISFY BUILDING CODES


The Current Building Codes Use the Terminology
Principal Direction without A Unique Definition

17.1. INTRODUCTION
Currently a three-dimensional dynamic analysis is required for a large number of
different types of structural systems that are constructed in Seismic Zones 2, 3 and 4
[1]. The lateral force requirements suggest several methods that can be used to
determine the distribution of seismic forces within a structure. However, these
guidelines are not unique and need further interpretations.
The major advantage of using the forces obtained from a dynamic analysis as the
basis for a structural design is that the vertical distribution of forces may be
significantly different from the forces obtained from an equivalent static load
analysis. Consequently, the use of dynamic analysis will produce structural designs
that are more earthquake resistant than structures designed using static loads.
For many years, approximate two-dimensional static load was acceptable as the
basis for seismic design in many geographical areas and for most types of structural
systems. During the past twenty years, due to the increasing availability of modern
digital computers, most engineers have had experience with the static load analysis
of three dimensional structures. However, few engineers, and the writers of the
current building code, have had experience with the three dimensional dynamic

STATIC AND DYNAMIC ANALYSIS

response analysis. Therefore, the interpretation of the dynamic analysis requirement


of the current code represents a new challenge to most structural engineers.
The current code allows the results obtained from a dynamic analysis to be
normalized so that the maximum dynamic base shear is equal to the base shear
obtained from a simple two-dimensional static load analysis. Most members of the
profession realize that there is no theoretical foundation for this approach.
However, for the purpose of selecting the magnitude of the dynamic loading that will
satisfy the code requirements, this approach can be accepted, in a modified form,
until a more rational method is adopted.
The calculation of the design base shears is simple and the variables are defined in
the code. It is of interest to note, however, that the basic magnitude of the seismic
loads has not changed significantly from previous codes. The major change is that
dynamic methods of analysis must be used in the principal directions of the
structure. The present code does not state how to define the principal directions for
a three dimensional structure of arbitrary geometric shape. Since the design base
shear can be different in each direction, this scaled spectra approach can produce
a different input motion for each direction, for both regular and irregular structures.
Therefore, the current code dynamic analysis approach can result in a structural
design which is relatively weak in one direction. The method of dynamic
analysis proposed in this chapter results in a structural design that has equal
resistance in all directions.
In addition, the maximum possible design base shear, which is defined by the present
code, is approximately 35 percent of the weight of the structure. For many
structures, it is less than 10 percent. It is generally recognized that this force level is
small when compared to measured earthquake forces. Therefore, the use of this
design base shear requires that substantial ductility be designed into the structure.
The definition of an irregular structure, the scaling of the dynamic base shears to the
static base shears for each direction, the application of accidental torsional loads and
the treatment of orthogonal loading effects are areas which are not clearly defined in
the current building code. The purpose of this section is to present one method of
three dimensional seismic analysis that will satisfy the Lateral Force Requirements
of the code. The method is based on the response spectral shapes defined in the code
and previously published and accepted computational procedures.

SEISMIC ANALYSIS MODELING

17.2. THREE DIMENSIONAL COMPUTER MODEL


Real and accidental torsional effects must be considered for all structures.
Therefore, all structures must be treated as three dimensional systems. Structures
with irregular plans, vertical setbacks or soft stories will cause no additional
problems if a realistic three dimensional computer model is created. This model
should be developed in the very early stages of design since it can be used for static
wind and vertical loads, as well as dynamic seismic loads.
Only structural elements with significant stiffness and ductility should be modeled.
Non-structural brittle components can be neglected. However, shearing, axial
deformations and non-center line dimensions can be considered in all members
without a significant increase in computational effort by most modern computer
programs. The rigid, in-plane approximation of floor systems has been shown to be
acceptable for most buildings. For the purpose of elastic dynamic analysis, gross
concrete sections, neglecting the stiffness of the steel, are normally used. A cracked
section mode should be used to check the final design.
The P-Delta effects should be included in all structural models. It has been shown in
Chapter 11 that these second order effects can be considered, without iteration, for
both static and dynamic loads. The effect of including P-Delta displacements in a
dynamic analysis results in a small increase in the period of all modes. In addition
to being more accurate, an additional advantage of automatically including P-Delta
effects is that the moment magnification factor for all members can be taken as unity
in all subsequent stress checks.
The mass of the structure can be estimated with a high degree of accuracy. The
major assumption required is to estimate the amount of live load to be included as
added mass. For certain types of structures it may be necessary to conduct several
analyses with different values of mass. The lumped mass approximation has proven
to be accurate. In the case of the rigid diaphragm approximation, the rotational
mass moment of inertia must be calculated.
The stiffness of the foundation region of most structures can be modeled by massless
structural elements. It is particularly important to model the stiffness of piles and
the rotational stiffness at the base of shear walls.

STATIC AND DYNAMIC ANALYSIS

The computer model for static loads only should be executed prior to conducting a
dynamic analysis.
Equilibrium can be checked and various modeling
approximations can be verified with simple static load patterns. The results of a
dynamic analysis are generally very complex and the forces obtained from a
response spectra analysis are always positive. Therefore, dynamic equilibrium is
almost impossible to check. However, it is relatively simple to check energy
balances in both linear and nonlinear analysis.

17.3. THREE DIMENSIONAL MODE SHAPES AND FREQUENCIES


The first step in the dynamic analysis of a structural model is the calculation of the
three dimensional mode shapes and natural frequencies of vibration. Within the past
several years, very efficient computational methods have been developed which have
greatly decreased the computational requirements associated with the calculation of
orthogonal shape functions as presented in Chapter 14. It has been demonstrated
that load-dependent Ritz vectors, which can be generated with a minimum of
numerical effort, produce more accurate results when used for a seismic dynamic
analysis than if the exact free-vibration mode shapes are used.
Therefore, a dynamic response spectra analysis can be conducted with
approximately twice the computer time requirements of a static load analysis. Since
systems with over 60,000 dynamic degrees-of-freedom can be solved within a few
hours on personal computers, there is not a significant increase in cost between a
static and a dynamic analysis. The major cost is the man hours required to
produce the three dimensional computer model that is necessary for a static or a
dynamic analysis.
In order to illustrate the dynamic properties of the three dimensional structure, the
mode shapes and frequencies are calculated for the irregular, eight story, 80 foot tall
building shown in Figure 17.1. This building is a concrete structure with several
hundred degrees-of-freedom. However, the three components of mass are lumped at
each of the eight floor levels. Therefore, only 24 three dimensional mode shapes are
possible.

SEISMIC ANALYSIS MODELING

Roof
8th

10 Typ.

7th
6th
5th
4th
3rd
2nd
Base

Figure 17.1. Example of Eight Story Irregular Building


Each three dimensional mode shape of a structure may have displacement
components in all directions. For the special case of a symmetrical structure, the
mode shapes are uncoupled and will have displacement in one direction only. Since
each mode can be considered to be a deflection due to a set of static loads, six base
reaction forces can be calculated for each mode shape. For the structure shown in
Figure 17.1, Table 17.1 summarizes the two base reactions and three overturning
moments associated with each mode shape. Since vertical mass has been neglected
there is no vertical reaction. The magnitudes of the forces and moments have no
meaning since the amplitude of a mode shape can be normalized to any value.
However, the relative values of the different components of the shears and moments
associated with each mode are of considerable value. The modes with a large
torsional component are highlighted in bold.

STATIC AND DYNAMIC ANALYSIS

Table 17.1. Three Dimensional Base Forces and Moments


MODE

PERIOD
Seconds

MODAL BASE SHEAR

MODAL OVERTURNING

REACTIONS

MOMENTS

X-DIR

Y-DIR

Angle Deg.

X-AXIS

Y-AXIS

Z-AXIS

.6315

.781

.624

38.64

-37.3

46.6

-18.9

.6034

-.624

.781

-51.37

-46.3

-37.0

38.3

.3501

.785

.620

38.30

-31.9

40.2

85.6

.1144

-.753

-.658

41.12

12.0

-13.7

7.2

.1135

.657

-.754

-48.89

13.6

11.9

-38.7

.0706

.989

.147

8.43

-33.5

51.9

2438.3

.0394

-.191

.982

-79.01

-10.4

-2.0

29.4

.0394

-.983

-.185

10.67

1.9

-10.4

26.9

.0242

.848

.530

32.01

-5.6

8.5

277.9

10

.0210

.739

.673

42.32

-5.3

5.8

-3.8

11

.0209

.672

-.740

-47.76

5.8

5.2

-39.0

12

.0130

-.579

.815

-54.63

-.8

-8.8

-1391.9

13

.0122

.683

.730

46.89

-4.4

4.1

-6.1

14

.0122

.730

-.683

-43.10

4.1

4.4

-40.2

15

.0087

-.132

-.991

82.40

5.2

-.7

-22.8

16

.0087

-.991

.135

-7.76

-.7

-5.2

30.8

17

.0074

-.724

-.690

43.64

4.0

-4.2

-252.4

18

.0063

-.745

-.667

41.86

3.1

-3.5

7.8

19

.0062

-.667

.745

-48.14

-3.5

-3.1

38.5

20

.0056

-.776

-.630

39.09

2.8

-3.4

54.1

21

.0055

-.630

.777

-50.96

-3.4

-2.8

38.6

22

.0052

.776

.631

39.15

-2.9

3.5

66.9

23

.0038

-.766

-.643

40.02

3.0

-3.6

-323.4

24

.0034

-.771

-.637

39.58

2.9

-3.5

-436.7

A careful examination of the directional properties of the three dimensional mode


shapes at the early stages of a preliminary design can give a structural engineer
additional information which can be used to improve the earthquake resistant design
of a structure. The current code defines an irregular structure as one which has a
certain geometric shape or in which stiffness and mass discontinuities exist. A far

SEISMIC ANALYSIS MODELING

more rational definition is that a regular structure is one in which there is a


minimum coupling between the lateral displacements and the torsional rotations for
the mode shapes associated with the lower frequencies of the system. Therefore, if
the model is modified and tuned by studying the three dimensional mode shapes
during the preliminary design phase, it may be possible to convert a geometrically
irregular structure to a dynamically regular structure from an earthquakeresistant design standpoint.
Table 17.2. Three Dimensional Participating Mass - (percent)
MODE

X-DIR

Y-DIR

Z-DIR

X-SUM

Y-SUM

Z-SUM

34.224

21.875

.000

34.224

21.875

.000

23.126

36.212

.000

57.350

58.087

.000

2.003

1.249

.000

59.354

59.336

.000

13.106

9.987

.000

72.460

69.323

.000

9.974

13.102

.000

82.434

82.425

.000

.002

.000

.000

82.436

82.425

.000

.293

17.770

.000

82.729

90.194

.000

7.726

.274

.000

90.455

90.469

.000

.039

.015

.000

90.494

90.484

.000

10

2.382

1.974

.000

92.876

92.458

.000

11

1.955

2.370

.000

94.831

94.828

.000

12

.000

.001

.000

94.831

94.829

.000

13

1.113

1.271

.000

95.945

96.100

.000

14

1.276

1.117

.000

97.220

97.217

.000

15

.028

1.556

.000

97.248

98.773

.000

16

1.555

.029

.000

98.803

98.802

.000

17

.011

.010

.000

98.814

98.812

.000

18

.503

.403

.000

99.316

99.215

.000

19

.405

.505

.000

99.722

99.720

.000

20

.102

.067

.000

99.824

99.787

.000

21

.111

.169

.000

99.935

99.957

.000

22

.062

.041

.000

99.997

99.998

.000

23

.003

.002

.000

100.000

100.000

.000

24

.001

.000

.000

100.000

100.000

.000

STATIC AND DYNAMIC ANALYSIS

For this building, it is of interest to note that the mode shapes, which tend to have
directions that are 90 degrees apart, have almost the same value for their period.
This is typical of three dimensional mode shapes for both regular and irregular
buildings. For regular symmetric structures, which have equal stiffness in all
directions, the periods associated with the lateral displacements will result in pairs of
identical periods. However, the directions associated with the pair of three
dimensional mode shapes are not mathematically unique. For identical periods, most
computer programs allow round-off errors to produce two mode shapes with
directions which differ by 90 degrees. Therefore, the SRSS method should not be
used to combine modal maximums in three dimensional dynamic analysis. The
CQC method eliminates problems associated with closely spaced periods.
For a response spectrum analysis, the current code states that at least 90 percent of
the participating mass of the structure must be included in the calculation of
response for each principal direction. Therefore, the number of modes to be
evaluated must satisfy this requirement. Most computer programs automatically
calculate the participating mass in all directions using the equations presented in
Chapter 13. This requirement can be easily satisfied using LDR vectors. For the
structure shown in Figure 17.1, the participating mass for each mode and for each
direction is shown in Table 17.2. For this building, only eight modes are required to
satisfy the 90 percent specification in both the x and y directions.

17.4. THREE DIMENSIONAL DYNAMIC ANALYSIS


It is possible to conduct a dynamic, time-history, response analysis by either the
mode superposition or step-by-step methods of analysis. However, a standard timehistory ground motion, for the purpose of design, has not been defined. Therefore,
most engineers use the response spectrum method of analysis as the basic approach.
The first step in a response spectrum analysis is the calculation of the three
dimensional mode shapes and frequencies as indicated in the previous section.
17.4.1. Dynamic Design Base Shear
For dynamic analysis, the 1994 UBC requires that the design base shear, V, is to
be evaluated from the following formula:
V = [ Z I C / RW ] W

(17.1)

SEISMIC ANALYSIS MODELING

Where
Z = Seismic zone factor given in Table 16-I.
I = Importance factor given in Table 16-K.
RW = Numerical coefficient given in Table 16-N or 16-P.
W = The total seismic weight of the structure.
C = Numerical coefficient (2.75 maximum value) determined from:
C = 1.25 S/ T2/3

(1-2)

Where
S = Site coefficient for soil characteristics given in Table 16-J.
T = Fundamental period of vibration (seconds).
The period, T, determined from the three dimensional computer model, can be used
for most cases. This is essentially Method B of the code.
Since the computer model often neglects nonstructural stiffness, the code requires
that Method A be used under certain conditions. Method A defines the period, T, as
follows:
T = Ct h3/4

(1-3)

where h is the height of the structure in feet and Ct is defined by the code for various
types of structural systems.
The Period calculated by Method B cannot be taken as more than 30% longer than
that computed using Method A in Seismic Zone 4 and more than 40% longer in
Seismic Zones 1, 2 and 3.
For a structure that is defined by the code as regular, the design base shear may be
reduced by an additional 10 percent. However, it must not be less than 80 percent
of the shear calculated using Method A. For an irregular structure this reduction
is not allowed.

10

STATIC AND DYNAMIC ANALYSIS

17.4.2. Definition of Principal Directions


A weakness in the current code is the lack of definition of the principal horizontal
directions for a general three dimensional structure. If each engineer is allowed to
select an arbitrary reference system, the dynamic base shear will not be unique
and each reference system could result in a different design. One solution to this
problem, that will result in a unique design base shear, is to use the direction of the
base shear associated with the fundamental mode of vibration as the definition of the
major principal direction for the structure. The minor principal direction will
be, by definition, ninety degrees from the major axis. This approach has some
rational basis since it is valid for regular structures. Therefore, this definition of the
principal directions will be used for the method of analysis presented in this chapter.
17.4.3. Directional and Orthogonal Effects
The required design seismic forces may come from any horizontal direction and, for
the purpose of design, they may be assumed to act non-concurrently in the direction
of each principal axis of the structure. In addition, for the purpose of member
design, the effects of seismic loading in two orthogonal directions may be combined
on a square-root-of-the-sum-of-the-squares (SRSS) basis. (Also, it is allowable to
design members for 100 percent of the seismic forces in one direction plus 30
percent of the forces produced by the loading in the other direction. We will not use
this approach in the procedure suggested here for reasons presented in Chapter 15.)
17.4.4. Basic Method of Seismic Analysis
In order to satisfy the current requirements, it is necessary to conduct two separate
spectrum analyses in the major and minor principal directions (as defined above).
Within each of these analyses, the Complete Quadratic Combination (CQC) method
is used to accurately account for modal interaction effects in the estimation of the
maximum response values. The spectra used in both of these analyses can be
obtained directly from the Normalized Response Spectra Shapes given by the
Uniform Building Code.
17.4.5. Scaling of Results
Each of these analyses will produce a base shear in the major principal direction. A
single value for the dynamic base shear is calculated by the SRSS method. Also,

SEISMIC ANALYSIS MODELING

11

a dynamic base shear can be calculated in the minor principal direction. The next
step is to scale the previously used spectra shapes by the ratio of design base
shear to the minimum value of the dynamic base shear. This approach is more
conservative than proposed by the current requirements, since only the scaling factor
that produces the largest response is used. However, this approach is far more
rational since it results in the same design earthquake in all directions.
17.4.6. Dynamic Displacements and Member Forces
The displacement and force distribution are calculated using the basic SRSS method
to combine the results from 100 percent of the scaled spectra applied in each
direction. If two analyses are conducted in any two orthogonal directions, in which
the CQC method is used to combine the modal maximums for each analysis, and the
results are combined by the SRSS method, exactly the same results will be obtained
regardless of the orientation of the orthogonal reference system. Therefore, the
direction of the base shear of the first mode defines a reference system for the
building.
If site-specific spectra are given, for which scaling is not required, any orthogonal
reference system can be used. In either case, only one computer run is necessary to
calculate all member forces to be used for design.
17.4.7. Torsional Effects
Possible torsional ground motion, the unpredictable distribution of live load mass
and the variations of structural properties are three reasons why both regular and
irregular structures must be designed for accidental torsional loads. Also, for a
regular structure lateral loads do not excite torsional modes. One method suggested
in the Code is to conduct several different dynamic analyses with the mass at
different locations. This approach is not practical since the basic dynamic
properties of the structure (and the dynamic base shears) would be different for each
analysis. In addition, the selection of the maximum member design forces would be
a monumental post-processing problem.
The current Code allows the use of pure static torsional loads to predict the
additional design forces caused by accidental torsion. The basic vertical distribution
of lateral static loads is given by the Code equations. The static torsional moment at

12

STATIC AND DYNAMIC ANALYSIS

any level is calculated by the multiplication of the static load at that level by 5
percent of the maximum dimension at that level. In this book it is recommended that
these pure torsional static loads, applied at the center of mass at each level, be used
as the basic approach to account for accidental torsional loads. This static torsional
load is treated as a separate load condition so that it can be appropriately combined
with the other static and dynamic loads.

17.5. NUMERICAL EXAMPLE


To illustrate the base-shear scaling method recommended here, a static seismic
analysis is conducted on the building shown in Figure 17.1. The eight-story building
has 10 feet story heights. The seismic dead load is 238.3 kips for the top four
stories and 363.9 kips for the lower four stories. For I = 1, Z = 0.4, S = 1.0, and
RW = 6.0, the evaluation of Equation 17.1 yields the design base forces given in
Table 17.3.Table 17.3. Static Design Base Forces Using The Uniform Building
Code

Period (sec)

Angle (deg)

Base Shear

Overturning
Moment

0.631

38.64

279.9

14,533

0.603

-51.36

281.2

14,979

The normalized response spectra shape for soil type 1, which is defined in the
Uniform Building Code, is used as the basic loading for the three dimensional
dynamic analyses. Using eight modes only and the SRSS method of combining
modal maxima, the base shears and overturning moments are summarized in Table
17.4 for various directions of loading.

13

SEISMIC ANALYSIS MODELING

Table 17.4. Dynamic Base Forces Using The SRSS Method


BASE SHEARS
Angle -deg

OVERTURNING MOMENTS

V1

V2

M1

M2

58.0

55.9

2982

3073

90

59.8

55.9

2983

3185

38.64

70.1

5.4

66

4135

-51.36

83.9

5.4

66

4500

The 1-axis is in the direction of the seismic input and the 2-axis is normal to the
direction of the loading. This example clearly illustrates the major weakness of the
SRSS method of modal combination. Unless the input is in the direction of the
fundamental mode shapes, a large base shear is developed normal to the direction of
the input and the dynamic base shear in the direction of the input is significantly
underestimated as illustrated in Chapter 15.
As indicated by Table 17.5, the CQC method of modal combination eliminates
problems associated with the SRSS method. Also, it clearly illustrates that the
directions of 38.64 and -51.36 degrees are a good definition of the principal
directions for this structure. Note that the directions of the base shears of the first
two modes differ by 90.00 degrees.
Table 17.5. Dynamic Base Forces Using The CQC Method
BASE SHEARS

OVERTURNING MOMENTS

Angle -deg

V1

V2

M1

M2

78.1

20.4

1202

4116

90

79.4

20.4

1202

4199

38.64

78.5

0.2

3.4

4145

-51.36

84.2

0.2

3.4

4503

Table 17.6 summarizes the scaled dynamic base forces to be used as the basis for
design by two different methods.

14

STATIC AND DYNAMIC ANALYSIS

Table 17.6 Normalized Base Forces In Principal Directions


38.64 Degrees

-51.36 Degrees

V
(kips)

M(ftkips)

V
(kips)

M(ft-kips)

Static Code Forces

279.9

14,533

281.2

14,979

Dynamic Design Forces


Scaled by Base Shear
279.9/78.5 = 3.57

279.9

14,732

299.2

16,004

For this case, the input spectra scale factor of 3.57 should be used for all directions
and is based on the fact that both the dynamic base shears and the dynamic
overturning moments must not be less than the static code forces. This approach is
clearly more conservative than the approach suggested by the current Uniform
Building Code. It is apparent that the use of different scale factors for a design
spectra in the two different directions, as allowed by the code, results in a design
that has a weak direction relative to the other principle direction.

17.6. DYNAMIC ANALYSIS METHOD SUMMARY


In this section, a dynamic analysis method is summarized that produces unique
design displacements and member forces which will satisfy the current Uniform
Building Code. It can be used for both regular and irregular structures. The major
steps in the approach are as follows:
1. A three dimensional computer model must be created in which all significant
structural elements are modeled. This model should be used in the early phases of
design since it can be used for both static and dynamic loads.
2. The three dimensional mode shapes should be repeatedly evaluated during the
design of the structure. The directional and torsional properties of the mode
shapes can be used to improve the design. A well-designed structure should have
a minimum amount of torsion in the mode shapes associated with the lower
frequencies of the structure.

SEISMIC ANALYSIS MODELING

15

3. The direction of the base reaction of the mode shape associated with the
fundamental frequency of the system is used to define the principal directions of
the three dimensional structure.
4. The design base shear is based on the longest period obtained from the
computer model, except when limited to 1.3 or 1.4 times the Method A calculated
period.
5. Using the CQC method, the dynamic base shears are calculated in each
principal direction due to 100 percent of the Normalized Spectra Shapes. Use the
minimum value of the base shear in the principal directions to produce one
scaled design spectra.
6. The dynamic displacements and member forces are calculated using the SRSS
value of 100 percent of the scaled design spectra applied non-concurrently in any
two orthogonal directions as presented in Chapter 15.
7. A pure torsion static load condition is produced using the suggested vertical
lateral load distribution defined in the code.
8. The member design forces are calculated using the following load combination
rule:
FDESIGN = FDEAD LOAD [ FDYNAMIC + | FTORSION | ] + FOTHER
The dynamic forces are always positive and the accidental torsional forces must
always increase the value of force. If vertical dynamic loads are to be considered, a
dead load factor can be applied.
One can justify many other methods of analyses that will satisfy the current code.
The approach presented in this chapter can be used directly with the computer
programs ETABS and SAP2000 with their steel and concrete post-processors.
Since these programs have very large capacities and operate on personal computers,
it is possible for a structural engineer to investigate a large number of different
designs very rapidly with a minimum expenditure of manpower and computer time.

16

STATIC AND DYNAMIC ANALYSIS

17.7. SUMMARY
After being associated with the three dimensional dynamic analysis and design of a
large number of structures during the past 40 years, the author would like to take
this opportunity to offer some constructive comments on the lateral load
requirements of the current code.
First: the use of the dynamic base shear as a significant indication of the
response of a structure may not be conservative. An examination of the modal base
shears and overturning moments in Tables 17.1 and 17.2 clearly indicates that base
shears associated with the shorter periods produce relatively small overturning
moments. Therefore, a dynamic analysis, which will contain higher mode response,
will always produce a larger dynamic base shear relative to the dynamic overturning
moment. Since the code allows all results to be scaled by the ratio of dynamic base
shear to the static design base shear, the dynamic overturning moments can be
significantly less than the results of a simple static code analysis. A scale factor
based on the ratio of the static design overturning moment to the dynamic
overturning moment would be far more logical. The static overturning moment can
be calculated by using the static vertical distribution of the design base shear which
is currently suggested in the code.
Second: for irregular structures, the use of the terminology period (or mode
shape) in the direction under consideration must be discontinued. The stiffness
and mass properties of the structure define the directions of all three dimensional
mode shapes. The term principal direction should not be used unless it is clearly
and uniquely defined.
Third: the scaling of the results of a dynamic analysis should be re-examined. The
use of site-dependent spectra is encouraged.
Finally: it is not necessary to distinguish between regular and irregular structures
when a three dimensional dynamic analysis is conducted. If an accurate three
dimensional computer model is created, the vertical and horizontal irregularities and
known eccentricities of stiffness and mass will cause the displacement and rotational
components of the mode shapes to be coupled. A three dimensional dynamic
analysis, based on these coupled mode shapes, will produce a far more complex
response with larger forces than the response of a regular structure. It is possible to

SEISMIC ANALYSIS MODELING

17

predict the dynamic force distribution in a very irregular structure with the same
degree of accuracy and reliability as the evaluation of the force distribution in a very
regular structure. Consequently, if the design of an irregular structure is based on a
realistic dynamic force distribution, there is no logical reason to expect that it will be
any less earthquake resistant than a regular structure which was designed using the
same dynamic loading. A reason why many irregular structures have a documented
record of poor performance during earthquakes is that their designs were often based
on approximate two dimensional static analyses.
One major advantage of the modeling method presented in this chapter is that one set
of dynamic design forces, including the effects of accidental torsion, is produced
with one computer run. Of greater significance, however, is the resulting structural
design has equal resistance to seismic motions from all possible directions.

17.8. REFERENCES
1. Recommended Lateral Force Requirements and Commentary, 1996 Sixth
Edition, Seismology Committee, Structural Engineers Association of California,
Tel. 916-427-3647.